Post-earnings-announcement drift anomaly: The role of operating and non-operating income in the Taiwanese stock market

Journal of Applied Finance & Banking, vol. 9, no. 4, 2019, 123-137
ISSN: 1792-6580 (print version), 1792-6599 (online)
Scienpress Ltd, 2019

Post-earnings-announcement drift anomaly: The
role of operating and non-operating income in the
Taiwanese stock market
Hsueh-Tien Lu1

Abstract
This paper examines the relationship between unexpected earnings components
(i.e., unexpected operating and non-operating income) and post-earningsannouncement drift to determine whether both components contribute to the
mispricing phenomenon. I find that both operating and non-operating income
surprises explain the market’s underweighting of earnings surprises. However, the
contribution of operating income surprises is significantly higher than
non-operating income surprises. While the mispricing of components appears to
be captured by post-earnings-announcement drift, the speed of price responses to
unexpected non-operating income is faster than for unexpected operating income.
Moreover, unexpected operating and non-operating income mispricing are distinct
mispricing phenomena, and a joint hedge portfolio trading strategy generates
excess abnormal returns when based only on an unexpected operating or
non-operating strategy.
JEL classification numbers: G14, M41
Keywords: Post-earnings-announcement drift, Operating income, Non-operating
income.

1 Introduction
Accounting principles indicate how to measure and when to report the effect of
economic events on the income statement. Reporting a firm’s profitability to

1

Department of Accounting, Chinese Culture University, Taiwan.

Article Info: Received: February 12, 2019. Revised: March 2, 2019
Published online: May 10, 2019


124

Hsueh-Tien Lu

stakeholders at periodic intervals is central to financial accounting. Reported
earnings alone may not communicate all the information in accounting data
needed to evaluate a firm’s profitability. The principles presume that the
classification scheme is informative enough about differences in the underlying
economic events and can represent a wide variety of economic events in order to
enhance the usefulness of an income statement. The accounting profession
requires that firms disaggregate reported earnings into operating income (captures
the results of the firm’s ongoing operations that will likely recur in the future) and
non-operating income (not part of ongoing operations and therefore less likely to
affect the firm’s performance in future periods).2 However, despite the significant
attention investors pay to firms’ income statements, most academic studies
contend that investors fail to fully incorporate the implications of earnings and its
components into stock prices in a timely fashion.
Post-earnings-announcement drift, first observed by Ball and Brown (1968) in the
United States, is the tendency for subsequent abnormal returns to move in the
direction of an earnings surprise for months after earnings are announced. This
predictability of abnormal stock returns after earnings-announcements has
attracted numerous and substantial research studies that found that
post-earnings-announcement drift is a robust phenomenon in the United States and

many other countries. Why the post-earnings-announcement drift anomaly has
been documented consistently and globally until now remains a puzzle for
researchers. One of the main explanations is that information processing biases
exist as a result of a delayed price response.3 Bernard and Thomas (1989, 1990)
indicates that immediate responses to earnings-announcements are not complete
and post-earnings-announcement drift is due to delayed reaction to the information
in earnings-announcements. Ball and Bartov (1996) show that investors underreact
to the magnitude of earnings surprises, and their underreaction is corrected at
future earnings-announcements.
The purpose of this paper is to investigate whether the patterns of investors
underreacting to the surprises are different across earnings, operating income, and
non-operating income. To some extent, the aggregated mispricing in response to
unexpected operating and non-operating income appears to be closely linked to
mispricing due to unexpected earnings. Since managers can use operating,
non-operating income or both to affect the sign (positive or negative) and
magnitude of an earnings surprise, the market may underreact to unexpected

2

Textbooks, practicing CPAs and financial analysts often suggest that certain components or
subtotals on the income statement provide more information than others regarding firm
profitability.
3
A large body of literature attempts to explain the drift; some explanations involve price
momentum (Chordia and Shivakumar, 2006), disclosure risk (Shin, 2005), arbitrage risk
(Mendenhall, 2004), information uncertainty (Francis et al., 2007), liquidity (Chordia, et al., 2009),
etc.


Post-earnings-announcement drift anomaly: The role of operating and …

125

operating and non-operating income occurring on the same time horizon, as well
as to unexpected earnings. A key question is whether the two components
represent a form of mispricing distinct from post-earnings-announcement drift.
Using a sample of 1,271 Taiwanese listed firms (21,787 firm-quarters from 2012
to 2016), my results provide evidence of significant, subsequent abnormal returns
associated with all of the quarterly unexpected earnings, operating and
non-operating income. More importantly, combining the unexpected earnings
strategy with unexpected operating or non-operating income strategies decreases
the magnitude of abnormal returns that can be earned, indicating that both the
mispricing of operating and non-operating income are part of the
post-earnings-announcement drift. Furthermore, my results show that the
contribution of operating income surprises to the earnings-based anomaly is
significantly higher than of non-operating income surprises. However, a joint
strategy of surprising operating and non-operating income increases the magnitude
of excess returns that can be earned. This result implies that investor
misperception of reported earnings disaggregated into operating and non-operating
income is more pronounced than of aggregated earnings. In addition, this paper
provides results that demonstrate larger price response delays for operating
income than for non-operating income. Nevertheless, price response speed is
similar for earnings and operating income, but faster price response for
non-operating income. Therefore, the results imply that stock prices do not reflect
operating and non-operating income in the same, timely fashion.
My findings contribute to the literature in two ways. First, this paper shows that
investors underreact to the information in operating and non-operating income
surprises and correct them at different speeds. This evidence complements the
delayed price response literature that reports different price response patterns
across operating and non-operating income. Second, my results support the notion

that subtotals on the income statement provide more incremental information than
earnings per share. Prior studies focus on the market reaction to different
components of earnings (e.g., Ohlson and Penman, 1992), and on the usefulness of
current financial reporting numbers for future earnings predictions (e.g., Finger,
1994). I add to these lines of research by suggesting that both operating and
non-operating income surprises are associated with post-earnings-announcement
drift.
The next section of this study is a brief review of previous research on pricing
earnings components. Section 3 describes the data and methodology. Section 4
outlines the tests and the results of my empirical findings. Section 5 provides a
conclusion.

2 Literature Review
Many studies focus on the information content of earnings components to examine


126

Hsueh-Tien Lu

the market reaction to different components of earnings. Gonedes (1975) indicates
that the market pricing of unusual earnings components is more influenced by the
sign (positive or negative) rather than the classification. Bowen (1981) shows that
investors put more value per dollar on operating components rather than on
non-operating ones. However, Bao and Bao (2004) show that the non-operating
income of Taiwanese firms has almost the same relevant value as their operating
income, suggesting that country-level institutional factors may affect the weight
placed by investors on earnings components. Strong and Walker (1993) show that
partitioning earnings into ordinary earnings, exceptional earnings, and
extraordinary items increases the association between abnormal returns and

earnings. Ohlson and Penman (1992) show that market reactions to earnings
components are divergent over short time horizons but are similar over longer
horizons. In sum, these studies suggest that the components provide different
information for market pricing. In this study, I test whether the surprised earnings
components contribute differently to the post-earnings-announcement drift
anomaly.
In addition, a large body of research focuses on examining market pricing based
on the different persistence properties of earnings components (e.g., Sloan, 1996;
Hui et al., 2016).4 These studies document that investors fail to distinguish the
different levels of persistence between earnings components leading to the
subsequent abnormal return due to market mispricing. The previous literature
proposes an explanation of investor fixation for the market mispricing of earnings
components (e.g., Xie, 2001; Harris et al., 2016). That is, investors fixate on
reported earnings and thus fail to incorporate information from the components of
current earnings. However, it is still unclear whether investor fixation on earnings
can fully explain the mispricing anomalies of earnings components (e.g., Dechow
et al., 2008; 2011). This paper adds to the literature by examining the contribution
of operating and non-operating income surprises on the mispricing of earnings
surprises.

3 Sample Selection and Methodology
3.1 Sample selection
I retrieved my sample data from the Taiwan Economic Journal (TEJ) and included
all firms publicly listed on the Taiwan Stock Exchange and Taipei Exchange. My
sample spans the period from 2012 to 2016, since annual financial reports must be
published after the end of each fiscal year and includes the four months before

4

Sloan (1996) studies the market mispricing on different levels of persistence between accruals

and cash flows. Hui et al. (2016) focus on pricing based on the persistence of industry-wide and
firm-specific earnings, cash flows, and accruals.


Post-earnings-announcement drift anomaly: The role of operating and …

127

2012 and the three months after the start of 2012.
The initial sample consists of all firm-quarters over the sample period. I exclude
the financial industry and firms with insufficient data to compute financial and
return variables. The final sample contains 21,787 firm-quarters for 1,271
Taiwanese listed firms.
3.2 Hedge portfolio approach
I first used a hedged portfolio approach to document that there is market
mispricing on unexpected earnings and its components (i.e., unexpected operating
and non-operating income, in the corresponding period of the following quarter).
The portfolio approach has the advantage that it addresses a potential, nonlinear
relationship between financial performance and stock returns (Fama, 1998;
Mitchell and Stafford, 2000; Levi, 2008).
When constructing a portfolio based on the magnitude of unexpected earnings,
operating income, or non-operating incomes, the hedged portfolio takes a long
position in the highest unexpected earnings component decile, and a short position
in the lowest unexpected earnings component decile; this generates positive future
returns. These results demonstrate the mispricing of unexpected earnings
components. I accumulated these returns over three different holding periods: (1,
5), (1, 21), and (1, second day before quarter t+1’s earnings-announcement). I
compared the mean size-adjusted returns for different holding horizons between
the hedge strategies of earnings components.5
3.3 Regression test

Next, I applied a regression approach that can be used to examine the association
between the unexpected earnings components and stock returns after controlling
for correlated, omitted variables for stock returns. The following two regressions
form the basis of the cross-sectionals:
BHARQi,t+1 (BHARN i,t+1) = α0 + α1UEi,t + α2SIZEi,t + α3BETAi,t + α4BTMi,t +
α5MOMi,t + ϵi,t+1
(1)
BHARQi,t+1 (BHARN i,t+1) = β0 +β1UOIi,t + β2UNOIi,t +β3SIZEi,t + β4BETAi,t +
β5BTMi,t + β6MOMi,t + ϵi,t+1
(2)
where BHARQ represents the size-adjusted, buy-and-hold returns for the period
beginning on the day after quarter t’s earnings-announcement and ending on the
5

In accordance with prior research (e.g. Bernard and Thomas 1990; Sloan 1996), I used
size-adjusted returns. In this paper, size-adjusted buy-and-hold return is the raw, buy-and-hold
return of the firm minus the mean buy-and-hold return of an equally weighted portfolio of firms
listed on the Taiwan Stock Exchange or Taipei Exchange in the same size decile over the same
holding period.


128

Hsueh-Tien Lu

second day before quarter t+1’s earnings-announcement date. BHARN is the 5-day
(BHAR5) or 21-day (BHAR21) size-adjusted, buy-and-hold returns after quarter t’s
earnings-announcement. Consistent with many prior studies (e.g., Livnat et al.,
2006), I estimated earnings surprised using a time-series, rolling, seasonal random
walk model. I defined the earnings surprise (UE) as earnings per share for quarter

t, minus earnings per share for quarter t-4, scaled by stock price per share at the
end of quarter t. Then, I included the unexpected earnings components variables
(UOI, and UNOI) to investigate the association between earnings components and
subsequent stock returns. This tells me something about the way earnings are
capitalized into prices. If the market correctly prices the information in historical
earnings, then the coefficients on earnings components variables should be
insignificant. Unexpected operating income (UOI) is calculated as operating
income per share for quarter t minus operating income per share for quarter t-4,
scaled by the price per share at the end of quarter t. Unexpected non-operating
income for quarter (UNOI) is calculated as non-operating income per share for
quarter t minus non-operating income per share for quarter t-4, scaled by the price
per share at the end of quarter t. Non-operating income is calculated as earnings
per share minus operating income.
These analyses control for a set of variables that prior literature shows to be
associated with subsequent stock returns. Specifically, I control for firm size
(SIZE), beta (BETA), book-to-market ratio (BTM), and momentum (MOM)
because prior studies have demonstrated that they are associated with future stock
returns (Carhart, 1997; Shivakumar, 2006).

4 Empirical Results
Table 1 provides statistics for the final sample based on the decile portfolios
formed by quarterly ranking firms on the magnitude of the earnings surprises.
Panel A reports the portfolio mean values for the magnitudes of unexpected
earnings (UE) and its two components (UOI and UNOI). The mean value of
unexpected operating income (non-operating income) falls from -0.050 (-0.031)
for the lowest unexpected earnings portfolio, to 0.050 (0.037) for the highest
unexpected earnings portfolio. The unexpected earnings trading strategy predicts
positive (negative) excess returns for firms in the most positive (negative) UE
decile. Thus, firms with large positive (negative) unexpected operating or
non-operating income that also belong to the most positive (negative) unexpected

earnings portfolio may tend to generate expected partial abnormal returns belong
to the unexpected earnings hedge strategy.


Post-earnings-announcement drift anomaly: The role of operating and …

129

Table 1: Mean values of variables by assigning deciles based on the magnitude of
unexpected earnings (N = 21,787)
Quarterly portfolio unexpected earnings ranking
Mean Lowest 2
3
4
5
6
7
8
9 Highest
Panel A: Components of unexpected earnings
0.001 -0.082 -0.022 -0.011 -0.005 -0.001 0.002 0.006 0.012 0.023 0.087
UE
UOI 0.000 -0.050 -0.018 -0.009 -0.004 -0.000 0.002 0.006 0.010 0.018 0.050
UNOI 0.000 -0.031 -0.004 -0.003 -0.002 -0.001 0.000 0.000 0.001 0.005 0.037
Panel B: Control variables
SIZE 6.542 6.405 6.466 6.571 6.626 6.680 6.669 6.606 6.541 6.464 6.395
BETA 0.761 0.786 0.781 0.748 0.734 0.737 0.748 0.751 0.775 0.777 0.768
BTM 1.044 0.905 1.018 1.050 1.098 1.140 1.113 1.092 1.071 1.032 0.924
MOM 0.088 -0.052 -0.027 0.013 0.036 0.060 0.085 0.115 0.153 0.206 0.294
Notes: UE is unexpected earnings for quarter t, which is calculated as earnings per share for

quarter t minus earnings per share for quarter t-4, scaled by the price per share at the end of quarter
t. UOI is unexpected operating income for quarter t, which is calculated as operating income per
share for quarter t minus operating income per share for quarter t-4, scaled by the price per share at
the end of quarter t. UNOI is unexpected non-operating income for quarter t, which is calculated as
non-operating income per share for quarter t minus non-operating income per share for quarter t-4,
scaled by the price per share at the end of quarter t. SIZE is the log of the market value at the end
of quarter t. BETA is the beta from the market model at the end of quarter t. BTM is the
book-to-market ratio at the end of quarter t. MOM is the stock return from twelve to two months
prior to the earnings-announcement month.

Panel B provides statistics on four risk proxies associated with future stock returns.
An inverted, U-shaped relationship in the portfolio indicates an extreme portfolio
containing smaller SIZE and lower BTM. A U-shaped relationship in the portfolio
indicates an extreme portfolio containing higher BETA. Those results show that
extreme portfolios are more risky. Across the unexpected earnings portfolios, the
mean values of the MOM range from -0.052 to 0.294. This reveals a positive
relationship between unexpected earnings and stock momentum.
Prior studies have documented that a positive relationship exists between
standardized unexpected earnings and future stock returns (e.g., Bernard and
Thomas, 1990). I sorted firm-quarters into deciles based on the levels of each
unexpected earnings components for the previous quarter. Then, I calculated mean
size-adjusted returns following the portfolio formation for each earnings
components. Table 2 compares the mean size-adjusted returns for different periods
following the prior year’s earnings-announcement for each unexpected earnings
components. I accumulated these returns over three holding periods: 5-days,
21-days, and one quarter.
Panel A of Table 2 provides the results for the unexpected earnings (UE) portfolio.
On average, a firm-quarter in the lowest (highest) unexpected earnings decile
experiences a downward (upward) price drift of -3.0 (5.0)% during the quarter

130

Hsueh-Tien Lu

after the prior quarter’s earnings-announcement. The quarterly hedged portfolio
return (taking a long position for the highest UE decile and a short position for the
lowest UE decile) is 8.0% (0.030 + 0.050). For the dissemination of current
earnings information regarding stock prices, the 5-day (21-day) hedged portfolio
returns are 4.0% (5.2%), which is 49.5% (64.9%) of the quarterly hedged portfolio
return. Panel B of Table 2 shows that the quarterly hedged portfolio returns of the
unexpected operating income (UOI) portfolio is 7.3% (0.028 + 0.045). In addition,
the 5-day (21-day) hedges portfolio returns are 3.2% (4.6%), which is 43.0%
(62.3%) of the quarterly hedged portfolio return. The unexpected operating
income (UOI) portfolio presents a slightly smaller hedged return and similar price
response speed compared to the unexpected earnings (UE) portfolio.
Table 2: Mean values across various portfolios based on the magnitude of
unexpected earnings (UE), unexpected operating income (UOI), and unexpected
non-operating income (UNOI) (N = 21,787)
Panel A: Mean returns across various portfolios based on the magnitude of UE
UE portfolio
N
UE
BHAR5 BHAR21 BHARQ
Lowest
2,169
-0.082
-0.019
-0.022
-0.030

2
2,181
-0.022
-0.012
-0.022
-0.029
3
2,177
-0.011
-0.008
-0.017
-0.025
4
2,180
-0.005
-0.005
-0.009
-0.011
5
2,181
-0.001
-0.002
-0.004
-0.006
6
2,176
0.002
0.001
0.000
0.002

7
2,178
0.006
0.004
0.005
0.010
8
2,179
0.012
0.006
0.012
0.023
9
2,179
0.023
0.014
0.021
0.029
Highest
2,187
0.087
0.021
0.030
0.050
Highest - Lowest
0.040
0.052
0.080
% of 1-Year Return
49.5%

64.9%
100.0%
Panel B: Mean returns across various portfolios based on the magnitude of UOI
UOI portfolio
N
UOI
BHAR5 BHAR21 BHARQ
Lowest
2,169
-0.068
-0.016
-0.020
-0.028
2
2,181
-0.021
-0.007
-0.013
-0.019
3
2,177
-0.010
-0.005
-0.013
-0.018
4
2,180
-0.005
-0.004
-0.008

-0.009
5
2,181
-0.001
-0.003
-0.005
-0.008
6
2,176
0.002
0.002
0.003
0.005
7
2,178
0.006
0.004
0.005
0.009
8
2,179
0.012
0.003
0.006
0.011
9
2,179
0.022
0.009
0.014

0.026
Highest
2,187
0.067
0.016
0.026
0.045
Highest - Lowest
0.032
0.046
0.073
% of 1-Year Return
43.0%
62.3%
100.0%


Post-earnings-announcement drift anomaly: The role of operating and …

131

Panel C: Mean returns across various portfolios based on the magnitude of UNOI
UNOI portfolio
N
UNOI
BHAR5 BHAR21 BHARQ
Lowest
2,169
-0.056
-0.007

-0.006
0.003
2
2,181
-0.013
-0.006
-0.007
-0.006
3
2,177
-0.006
0.001
-0.001
0.003
4
2,180
-0.003
0.000
-0.001
-0.002
5
2,181
-0.001
0.001
0.001
0.004
6
2,176
0.001
-0.001

-0.004
-0.007
7
2,178
0.003
0.000
0.002
0.001
8
2,179
0.006
0.001
-0.001
-0.002
9
2,179
0.012
0.001
0.000
0.000
Highest
2,187
0.061
0.009
0.012
0.020
Highest - Lowest
0.015
0.019
0.017

% of 1-Year Return
90.6%
110.3%
100.0%
Notes: BHAR5 (BHAR21) is the 5-day (21-day), size-adjusted, buy-and-hold returns after quarter
t’s earnings-announcement. BHARQ is the size-adjusted buy-and-hold return for the period
beginning on the day after quarter t’s earnings-announcement and ending on the second day before
quarter t+1’s earnings-announcement date. See the Table 1 for definitions of the other variables.

Panel C of Table 2 shows that the quarterly hedged portfolio returns of the
unexpected non-operating income (UNOI) portfolio is 1.7% (-0.003 + 0.020). The
5-day (21-day) hedged portfolio returns are 1.5% (1.9%), which is 90.6% (110.3%)
of the quarterly hedged portfolio returns. Compared to the unexpected earnings
(UE) portfolio, the unexpected non-operating income (UNOI) portfolio shows a
significantly smaller hedge return, but a faster price response. In sum, the delayed
market response is smaller and faster for unexpected non-operating income (UNOI)
than for unexpected operating income (UOI).
So far the unexpected earnings, operating and non-operating income strategies
have been independently examined. If the market’s mispricing of unexpected
operating or non-operating income is part of the post-earnings-announcement drift,
then it should be possible to form trading strategies that capitalize on an
unexpected earnings strategy with operating or non-operating income strategies
that yield smaller hedge returns than the unexpected earnings strategy in Panel A
of Table 2.
Table 3 shows a contingency table of abnormal returns earned from portfolios
constructed by grouping together firms according to all of the unexpected earnings,
operating and non-operating income. The numbers of firm-quarters in each cell are
reported in parentheses. To simplify, quintiles 2-4 have been condensed into a
single cell, while the extreme quintiles (1 and 5) are presented separately. Panel A
of Table 3 presents the results of a joint strategy formed by unexpected earnings

(UE) and unexpected operating income (UOI). A hedged portfolio strategy formed
by taking a long position in UE5/UOI5 firms and a short position in UE1/UOI1
firms will earn an abnormal return of 7.7% (0.045+0.032) for one quarter, slightly


132

Hsueh-Tien Lu

smaller than the unexpected earnings strategy (8.0%). Panel B of Table 3 presents
the results of a joint strategy constructed by unexpected earnings (UE) and
unexpected non-operating income (UNOI). A hedged portfolio strategy formed by
the extreme quintiles will earn an abnormal return of 6.3% (0.038+0.025) for one
quarter, smaller than the unexpected earnings strategy (8.0%). These results imply
that the price response to unexpected earnings has incorporated the information of
unexpected operating and non-operating income. In addition, both unexpected
operating and non-operating income could result in the post-earningsannouncement drift phenomenon.
Table 3: Double portfolio sorting (N = 21,787)
Panel A: Double portfolio sorting based upon unexpected earnings (UE) and
unexpected operating income (UOI)
UE quintile
UOI1
UOI2-4
UOI5
UNOI1
-0.014
0.044
-0.024
-0.032
(3004)

(1124)
(222)
(4350)
UNOI2-4
-0.025
-0.002
0.022
-0.002
UOI
(1126)
(10841)
(1104)
(13071)
quintile
UNOI5
-0.020
0.019
0.035
0.045
(220)
(1106)
(3040)
(4366)
-0.030
-0.001
0.040
(4350)
(13071)
(4366)
Panel B: Double portfolio sorting based upon unexpected earnings (UE) and

unexpected non-operating income (UNOI)
UE quintile
UOI1
UOI2-4
UOI5
UNOI1
-0.001
0.049
-0.002
-0.025
(1768)
(1782)
(800)
(4350)
UNOI2-4
-0.033
-0.001
0.037
-0.001
UNOI
quintile
(1730)
(9614)
(1727)
(13071)
UNOI5
-0.031
0.000
0.010
0.038

(852)
(1675)
(1839)
(4366)
-0.030
-0.001
0.040
(4350)
(13071)
(4366)


Post-earnings-announcement drift anomaly: The role of operating and …

133

Panel C: Double portfolio sorting based upon unexpected operating income (UOI)
and unexpected non-operating income (UNOI)
UOI quintile
UOI1
UOI2-4
UOI5
UNOI1
-0.019
0.027
-0.002
-0.031
(607)
(1938)
(1805)

(4350)
UNOI2-4
-0.032
-0.002
0.037
-0.001
UNOI
quintile
(1911)
(9295)
(1865)
(13071)
UNOI5
-0.013
0.015
0.010
0.055
(1832)
(1838)
(696)
(4366)
-0.024
-0.002
0.035
(4350)
(13071)
(4366)
Notes: BHAR5 (BHAR21) is the 5-day (21-day), size-adjusted, buy-and-hold returns after quarter
t’s earnings-announcement. BHARQ is the size-adjusted buy-and-hold return for the period
beginning on the day after quarter t’s earnings-announcement and ending on the second day before

quarter t+1’s earnings-announcement date. See Table 1 for definitions of the other variables. The
number of observations per cell is reported in parentheses.

Furthermore, I constructed a contingency table of abnormal returns earned from
portfolios by grouping firms according to unexpected operating and non-operating
income in Panel C of Table 3. In this matrix, a hedged portfolio strategy formed by
the extreme quintiles will earn an abnormal return of 8.6% (0.055+0.031) for one
quarter, larger than that of individual an unexpected operating income strategy
(7.3%) or an unexpected non-operating income strategy (1.7%). This result
confirms that the market’s mispricing of unexpected operating and non-operating
income is distinct from each other.
Table 4 reports the results of pooled cross-sectional regressions (Panel A) and
decile rank regressions (Panel B). 6 I regressed future returns on explanatory
variables that might affect the magnitude of the delayed price response. The
dependent variables are the 5-day (BHAR5), 21-day (BHAR21), and single-quarter
(BHARQ), size-adjusted, buy-and-hold returns following the prior quarter’s
earnings-announcement date. Consistent with the post-earnings-announcement
drift literature, the coefficients on UE are positive and significant. Investors
underestimate the standardized earnings surprise for the subsequent quarter’s
earnings, resulting in higher future returns for firms with higher unexpected
earnings.

6

Decile rank regression alleviates problems associated with extreme values that are not
representative of the population or are measured with error.


134

Hsueh-Tien Lu

Table 4: Regression results of abnormal returns across various holding periods (N = 21,787)

Panel A: Actual values
Variable
BHAR5
Con.
UE
UOI
UNOI
SIZE

-0.012
(-2.93)***
0.219
(25.05)***

BHAR21

BHARQ

BHAR5

BHAR21

BHARQ

-0.026
(-3.65)***

0.313
(19.70)***

-0.053
(-4.42)***
0.472
(17.67)***

-0.012
(-3.04)***

-0.027
(-3.74)***

-0.054
(-4.49)***

0.257
(23.12)***
0.214
(15.93)***
0.001
(2.05)**
-0.002
(-2.18)**
0.005
(6.88)***
0.003
(2.87)***
0.033

0.043***

0.386
(19.14)***
0.291
(11.95)***
0.002
(2.19)**
0.004
(2.96)***
0.007
(5.49)***
0.000
(0.28)
0.021
0.095***

0.593
(17.49)***
0.396
(9.67)***
0.005
(2.93)***
-0.001
(-0.62)
0.018
(9.13)***
0.002
(0.73)
0.020

0.197***

0.001
0.002
0.005
(1.96)**
(2.12)**
(2.87)***
BETA
-0.002
0.004
-0.002
**
***
(-2.32)
(2.82)
(-0.76)
BTM
0.005
0.007
0.018
***
***
(6.84)
(5.47)
(9.10)***
MOM
0.003
0.001
0.003

***
(3.34)
(0.83)
(1.29)
Adj. R2
0.032
0.020
0.018
Difference in sensitivity between UOI and UNOI
Panel B: Decile ranking values
Variable
BHAR5
BHAR21
BHARQ
BHAR5
BHAR21 BHARQ
Con.
-0.025
-0.039
-0.059
-0.037
-0.056
-0.078
(-14.84)*** (-12.35)*** (-10.50)*** (-18.67)*** (-15.28)*** (-12.19)***
UE
0.039
0.059
0.088
(29.35)***
(24.06)***

(20.45)***
UOI
0.035
0.056
0.085
(24.65)*** (21.37)*** (18.49)***
UNOI
0.023
0.032
0.037
(16.77)*** (12.52)*** (8.48)***
SIZE
0.001
0.003
0.001
0.001
0.003
0.001
(0.90)
(1.13)
(0.14)
(0.95)
(1.20)
(0.20)
BETA
-0.002
0.006
0.001
-0.002
0.006

0.001
**
**
(-1.53)
(2.43)
(0.22)
(-1.43)
(2.47)
(0.24)
BTM
0.006
0.006
0.021
0.007
0.006
0.021
***
**
***
***
**
(4.60)
(2.46)
(4.68)
(4.72)
(2.50)
(4.67)***
MOM
0.002
-0.004

-0.002
0.003
-0.002
-0.001
**
(1.27)
(-1.48)
(-0.41)
(2.38)
(-0.94)
(-0.16)
Adj. R2
0.042
0.027
0.021
0.035
0.023
0.018
***
***
Difference in sensitivity between UOI and UNOI
0.012
0.024
0.048***
Notes: BHAR5 (BHAR21) is the 5-day (21-day), size-adjusted, buy-and-hold returns after quarter
t’s earnings-announcement. BHARQ is the size-adjusted buy-and-hold return for the period


Post-earnings-announcement drift anomaly: The role of operating and …

135

beginning on the day after quarter t’s earnings-announcement and ending on the second day before
quarter t+1’s earnings-announcement date. In Panel A, all variables are winsorized at the 1% and
99% levels. In Panel B, the decile ranks for each variable (ranked 1, 2,…, 10) are calculated for
each sample quarter. ***, **, * represent significance at the 1%, 5%, and 10% levels, respectively.
Two-tailed t-values are reported in parentheses.

Investors may have different reactions to different components of unexpected
earnings. Thus, I tested whether stock prices equally reflect both the unexpected
operating and non-operating components of the one-quarter-ahead unexpected
earnings. The regression coefficients on both the unexpected operating income
(UOI) and the unexpected non-operating income (UNOI) are positive and
significant, so the market underestimates both operating and non-operating income
surprises. However, the coefficients on UOI are much higher than the coefficients
on UNOI, suggesting that the market appears to underprice unexpected operating
income to a greater extent than it underprices unexpected non-operating income.
Together, these results imply that both the unexpected operating and non-operating
income contribute to the post-earnings-announcement drift and the unexpected
operating income plays a more significant role in the market anomaly than
unexpected non-operating income.

5 Conclusions
Previous studies argue that the market systematically underestimates the
persistence of earnings surprises resulting from the post-earnings-announcement
drift anomaly (e.g., Ball and Bartov, 1996). This paper examined the relationship
between unexpected earnings components (unexpected operating and
non-operating income) and post-earnings-announcement drift to see if both
components contribute to the mispricing phenomenon. Specifically, if earnings
surprises are associated with the permanent components of earnings, the market

may only underreact to operating income surprises rather than to non-operating
income surprises, the transitory components of earnings. The evidence provided in
this paper shows that both the operating and non-operating income surprises are
associated with the post-earnings-announcement drift. However, the contribution
of operating income surprises is significantly higher than non-operating income
surprises.
While the both the operating and non-operating income surprises explain the
market’s underweighting of earnings surprises, the speed of price responses to
non-operating income is faster than to operating income. For instance, my results
show that the markets reflect 90.6% (110.3%) subsequent quarter abnormal
returns in a 5-day (21-day) window for unexpected non-operating income
compared to 43.0% (62.3%) subsequent quarter abnormal returns in a 5-day
(21-day) window for unexpected operating income. Furthermore, this paper
provides evidence that unexpected operating and non-operating income appear to


136

Hsueh-Tien Lu

capture different mispricing phenomenon by combining the operating-based
strategy with the non-operating-based strategy. This joint strategy of operating and
non-operating income surprises increases the magnitude of excess returns that can
be earned by individual strategies.

References
[1] Ball, R. and Bartov, E., How Naive is the Stock Market's Use of Earnings
Information? Journal of Accounting and Economics, 21(3), 1996, pp.
319-337.
[2] Ball, R. and Brown, P., An Empirical Evaluation of Accounting Income

Numbers. Journal of Accounting Research, 6(2), 1968, pp. 159-178.
[3] Bao, B.H., Bao, D.H., Value Relevance of Operating Income versus
Non-operating Income in the Taiwan Stock Exchange. Advances in
International Accounting, 17(1), 2004, pp. 103-117.
[4] Bernard, V.L., and Thomas, J.K., Post-Earnings-Announcement Drift:
Delayed Price Response or Risk Premium? Journal of Accounting Research,
27, 1989, pp. 1-36.
[5] Bernard, V. L., and Thomas, J. K., Evidence that Stock Prices Do Not Fully
Reflect the Implications of Current Earnings for Future Earnings. Journal of
Accounting and Economics, 13(4), 1990, pp. 305-340.
[6] Bowen, R., The Valuation of Earnings Components in the Electric Utility
Industry. The Accounting Review, 56 (1), 1981, pp. 1-22.
[7] Carhart, M.M., On Persistence in Mutual Fund Performance. Journal of
Finance, 52(1), 1997, pp. 57-82.
[8] Chordia, T., and Shivakumar, L., Earnings and Price Momentum. Journal of
Financial Economics, 80(3), 2006, pp. 627-656.
[9] Chordia, T., Goyal, A., Sadka, G., Sadka, R., and Shivakumar, L., Liquidity
and the Post-Earnings-Announcement Drift. Financial Analysts Journal,
65(4), 2009, pp. 18-32.
[10] Dechow, P.M., Ge, W., Larson, C.R., and Sloan, R.G., Predicting Material
Accounting Misstatements. Contemporary Accounting Research, 28(1), 2011,
pp. 17-82.
[11] Dechow, P.M., Richardson, S.A., Sloan, R.G., The Persistence and Pricing of
the Cash Component of Earnings. Journal of Accounting Research. 46(3),
2008, pp. 537-566.
[12] Fama, E.F., Market Efficiency, Long-Term Returns, and Behavioral Finance.
Journal of Financial Economics, 49(3), 1998, pp. 283-306.
[13] Finger, C., The Ability of Earnings to Predict Future Earnings and Cash Flow.
Journal of Accounting Research, 32(2), 1994, pp. 210-223.
[14] Francis, J., LaFond, R., Olsson, P., and Schipper, K., Information Uncertainty

and the Post-Earnings-Announcement Drift. Journal of Business Finance and
Accounting, 34(3-4), 2007, pp. 403-433.


Post-earnings-announcement drift anomaly: The role of operating and …

137

[15] Gonedes, N.J., Efficient Capital Markets and External Accounting. The
Accounting Review, 47(1), 1972, pp. 11-21.
[16] Hui, K.W., Nelson, K.K., and Yeung, P.E., On the Persistence and Pricing of
Industry-Wide and Firm-Specific Earnings, Cash Flows, and Accruals.
Journal of Accounting and Economics. 61(1), 2016, pp. 185-202.
[17] Levi, S., Voluntary Disclosure of Accruals in Earnings Press Releases and the
Pricing of Accruals. Review of Accounting Studies, 13(1), 2008, pp. 1-21.
[18] Mendenhall, R., Arbitrage Risk and Post-Earnings-Announcement Drift.
Journal of Business, 77(4), 2004, pp. 875-894.
[19] Mitchell, M.L., and Stafford, E., Managerial Decisions and Long‐Term Stock
Price Performance. Journal of Business, 73(3), 2000, pp. 287-329.
[20] Ohlson, J.A., and Penman, S.H., Disaggregated Accounting Data as
explanatory Variables for Returns. Journal of Accounting, Auditing and
Finance, 7(4), 1992, pp. 553-573.
[21] Shin, H.S., Disclosure Risk and Price Drift. Journal of Accounting Research,
44(2), 2005, pp. 351-379.
[22] Shivakumar, L., Accruals, Cash Flows and the Post‐Earnings‐Announcement
Drift. Journal of Business Finance and Accounting, 33(1‐2), 2006, pp. 1-25.
[23] Sloan, R.G., Do Stock Prices Fully Reflect Information in Accruals and Cash
Flows About Future Earnings? The Accounting Review, 71(3), 1996, pp.
289–315.
[24] Strong, N., and M. Walker., The Explanatory Power of Earnings for Stock

Returns. The Accounting Review, 68 (2), 1993, pp. 385-399.
[25] Xie, H., The Mispricing of Abnormal Accruals. The Accounting Review,
76(3), 2001, pp. 357-373.